Hi there, I'm Dr. Ewe Win Eng, Ph.D 👋

⚛️ Data Scientist | Machine Learning Engineer | Commercial AI Systems

I am a Glasgow-based Data Scientist bridging the gap between deep mathematical rigor and commercial AI product development. I specialize in Physics-Informed Machine Learning, utilizing TensorFlow and LSTMs , and advanced statistical models to optimize complex systems and generate direct business value.

• 👨‍💻 Role: Data Scientist, Artificial Intelligence / Machine Learning Engineer.
• 🚀 Business Impact: I build production-ready predictive models and simulations that optimize efficiency, automate valuation pipelines, and reduce operational costs.
• 🧠 Top Skills: 🐍 Python, Deep Learning (TensorFlow/PyTorch), 📉 Time-Series Forecasting, and Automated ETL Pipelines.
• ⚡ Core Expertise: Renewable Energy Systems, Thermodynamics, Systems Modelling, and Predictive Analytics.
• 🔭 Currently working on: Integrating Deep Learning with Subsurface Thermal Energy Storage (STEaM) simulations to predict long-term thermal behavior.
• 🤝 Looking to collaborate on: AI-driven energy decarbonization projects and predictive maintenance models.
• 🇬🇧 Status: Endorsed by the UK Government as an exceptional talent. UK Global Talent Visa Holder – I can work for any employer immediately without sponsorship.

---

🛠️ Technical Skills

Category	Stack
Cloud Platforms
Databases & Data Processing
Dev Tools
Languages
Machine Learning
Visualization
Research Domain	Time-Series Forecasting Thermodynamics Renewable Energy Deep Learning Optimization A/B Testing Regression Classification

---

📊 Featured Projects (Portfolio)

1. STEaM: Mine Shaft Thermal Energy Storage Simulation

Industry Application: Renewable Energy Grid Balancing & Techno-Economic Modelling

• The Challenge: De-risking the conversion of legacy mine shafts into GigaWatt-hour thermal storage required complex modeling to validate feasibility and grid balancing potential.
• The Solution: Engineered a Python-based finite volume simulation engine (shaftstore_1d_0i.py) to model thermodynamic stratification and heat diffusion, integrating control logic for Heat Pump and CHP systems.
• The Impact: Generated critical Levelized Cost of Heat (LCOH) and COP metrics, providing the validation needed to repurpose industrial liabilities into renewable energy assets.
• Stack: Python NumPy Pandas SciPy Matplotlib Finite Volume Method

2. PVT Optimization: Genetic Algorithms

Industry Application: System Efficiency Improvement

• The Challenge: Solar collectors were underperforming due to static configuration parameters.
• The Solution: Wrote custom Genetic Algorithms (Optimization) to cycle through thousands of design variables.
• The Impact: Identified a configuration that increased energy capture by 30%.
• Stack: MATLAB Optimization Data Visualization

3. London Electrical Consumption Predictor

Industry Application: Energy Sector & Predictive Analytics

• The Challenge: A fictional London energy company needed to accurately forecast the daily total electrical consumption of customers across different boroughs.
• The Solution: Developed a deep learning pipeline that joined historical hourly energy usage with daily weather metrics, leveraging advanced feature engineering and a 7-stage data science lifecycle.
• The Impact: Delivered highly accurate predictive insights for energy demand, translating complex mathematical results into clear, data-driven business recommendations.
• Stack: Python Scikit-Learn Data Analysis Deep Learning Pandas Data Visualisation Tensorflow

4. Diamond Price Prediction: Deep ResNet-MLP

Industry Application: Automated Valuation & Pricing Engines

• The Challenge: Traditional linear models failed to capture complex non-linear interactions between categorical attributes (cut, clarity) and price for accurate valuation.
• The Solution: Engineered a custom ResNet-MLP (Deep Learning) architecture using TensorFlow/Keras, implementing residual skip connections and Log-Norm target engineering to stabilize gradients.
• The Impact: Delivered a production-ready pipeline capable of real-time price inference, targeting an accuracy of R² > 0.95.
• Stack: TensorFlow Keras Pandas Scikit-Learn ResNet

5. NHS Covid-19 Demand Modelling

Industry Application: Healthcare Analytics & Resource Planning

• The Challenge: The Scottish Government needed rapid projections of ICU bed usage.
• The Solution: Applied statistical modelling to patient intake data to forecast demand spikes.
• The Impact: Directly supported public health resource planning during a critical crisis.
• Stack: Python Scikit-Learn Data Analysis